Forecasting GDP Growth from Outer Space
| Date | 01 August 2020 |
| Published date | 01 August 2020 |
| DOI | http://doi.org/10.1111/obes.12361 |
| Author | Jaqueson K. Galimberti |
697
©2020 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd.
OXFORD BULLETIN OF ECONOMICSAND STATISTICS, 82, 4 (2020) 0305–9049
doi: 10.1111/obes.12361
Forecasting GDP Growth from Outer Space*
Jaqueson K. Galimberti
Auckland University of Technology, B-31 Economics, Private mail bag 92006, Auckland,
1142, New Zealand. (e-mail: jaqueson.galimberti@aut.ac.nz, https://sites.google.com/site/
jkgeconoeng/)
Abstract
We evaluate the usefulness of satellite-based data on night-time lights for forecasting
GDP growth across a global sample of countries, proposing innovative location-based
indicators to extract new predictive information from the lights data. Our findings are
generally favourable to the use of night lights data to improve the accuracy of model-
based forecasts. We also find a substantial degree of heterogeneity across countries in
the relationship between lights and economic activity: individually estimated models tend
to outperform panel specifications. Key factors underlying the night lights performance
include the country’s size and income level, logistics infrastructure, and the quality of
national statistics.
I. Introduction
Forecasts of economic activity are crucial to the decision-making process of policymakers
and market participants in general. A premise for informed economic decisions is to have
a proper expectation of the future state of the market of interest. In practice, the decision-
maker is then continuously faced with an intricate forecasting challenge of finding leading
indicators for the variables that are relevant to her/his business. In this paper, we propose
and evaluate the usage of satellite-based data on night-time lights for the prediction of
GDP growth across a global sample of countries. Our main contribution is the design of
innovative measures for the extraction of predictive signals of macroeconomic activity
from the richness of information provided by the night lights dataset.
JEL Classification numbers: C55, C82, E01, E37, R12.
*The research summarized in this paper was presented previously to the audience of several meetings: the
36th International Symposium on Forecasting, Santander, 2016, the 23rd International Conference Computing in
Economics and Finance, NewYork, 2017, the 70th European Meeting of the Econometric Society, Lisbon, 2017, the
SECO/SNB/KOFMeeting, Zurich, 2017, the Econometrics and Business Statistics Department at Monash University
Melbourne, 2017, the 1st ViennaWorkshop on Economic ForecastingVienna, 2018, the JRC-CAS Workshop on Big
Data and Macroeconomic Forecasting, Ispra, 2018, and the 34th Conference of the Centre for International Research
on Economic Tendency Surveys, Rio de Janeiro, 2018. I thank the comments and suggestions made by several par-
ticipants of these meetings, and by my former colleagues at the Chair of Applied Macroeconomics and KOF/ETH
Zurich. Financial support from the SAS-IIF grant is gratefully acknowledged; I also thank the corresponding grant
committee for the feedback. Finally,I also thank the editor, Professor Jonathan Temple,and two anonymous referees,
for their very constructive comments that greatly contributed to improve this paper.Any remaining errors are my
own.
698 Bulletin
The night lights data consist of gridded observations of light intensities captured across
the globe during night time. In order to evaluate the usefulness of these data for economic
forecasting we construct Sum of Lights (SoL) measures, aggregating the intensities of
lights observed within the borders of each country. One key innovation in this paper is
the development of alternative location-based SoL indicators, designed to focus on the
lights emitted from selected areas instead of the entire country’s territory; particularly,
we propose focusing on areas showing (significant) positive/negative correlations with the
country’s history of GDP growth rates. As our results reveal, a substantial portion (≈85%
on average) of the lights signals observed over a country’s territory are not significantly
correlated with the country’s aggregate production, and would therefore only add noise to
an indiscriminate aggregation of the country’s lights. We show how a proper classification
of these geo-located signals can lead to a substantial improvement of the accuracy of the
night lights-based forecasts.
Another important contribution of this paper is an examination of alternative assump-
tions on the cross-country specification of the relationship between light emissions and
economic growth. Namely, we question the common practice of assuming that this rela-
tionship is homogeneous across countries, and show that there are substantial accuracy
improvements to be achievedas well by allowing for individual country or partially pooled
specifications. We also find that the heterogeneity of performances of the night lights-based
forecasts can be associated with some country-specific factors, such as the country’s size,
income level, expenditure and production composition, logisticsinfrastr ucture, and quality
of national statistics.
Motivation and relation to literature
The use of night lights data has been prominent in the recent economic literature, with
applications that range from the geographical mapping of economic activity (Sutton and
Costanza, 2002; Doll, Muller and Morley, 2006; Ghosh et al., 2010) to regional develop-
ment analysis (Michalopoulos and Papaioannou, 2013a,b), to the evaluationof the accuracy
of national income accounts (Chen and Nordhaus, 2011; Henderson, Storeygard and Weil,
2012; Nordhaus and Chen, 2015; Pinkovskiy and Sala-i Martin, 2016); see also Donaldson
and Storeygard (2016) for a more general review of applications of satellite-based data in
economics. In order to construct comparative measures of living standards across countries
and regions, these studies have focused either on time-averaged relationships, hence taking
advantage mainly of the geographical dimension of the luminosity data variability, or on
the contemporaneous relationship between light emissions and economic activity.
Here, in contrast, we focus on the (lagged) time variations in the intensity of night
lights within a country, evaluating their usefulness to improve the accuracy of forecasts
of economic activity.To the best of our knowledge this is the first application of the night
lights data to economic forecasting at a global scale, and this gap in the literature seems to
be associated with the difficulty in providing a rationale for seeing night lights as a leading
indicator for the aggregate economy.1Thus, in order to challenge this view, we motivate
1Focusingon the case of China, Zhao et al. (2017) used the lights data for forecasting economic activity at different
levels of regional aggregation.
©2020 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
To continue reading
Request your trial